A flow measurement is never the product of a single perfect instrument. It is the combined result of a meter, a transmitter, a temperature element, a pressure sensor, and a flow computer, each carrying its own small error. Total probable error, or TPE, is the honest, statistically correct way to roll those separate errors into one number that describes how far the reported flow could realistically be from the true flow.
Total Probable Error (TPE) in one line: Total probable error is the combined uncertainty of a flow measurement, calculated by taking the square root of the sum of the squares of each independent error source rather than simply adding them. This root-sum-square method reflects the low probability that all errors reach their maximum in the same direction at once, so TPE gives a realistic worst-likely figure instead of an inflated absolute worst case.
The instinct is to add every error together. If the meter is good to one part, the pressure transmitter to another, and the temperature element to a third, surely the total is the arithmetic sum. That number, however, describes a situation that almost never happens: every device simultaneously wrong by its full rated amount, and all in the same direction. For independent errors that behave like random draws, the probability of that alignment is vanishingly small, so a straight sum badly overstates the realistic error.
Root sum square, or RSS, captures the more likely outcome. You square each individual error, add the squares, then take the square root. Because squaring rewards the large terms and shrinks the small ones, the combined result is dominated by the biggest contributors and barely moved by the little ones. Two half-percent errors do not combine to one percent by RSS; they combine to about seven tenths of a percent, because it is unlikely both peak together.
RSS is valid only when the error sources are genuinely independent of one another. If two contributors share a common cause, such as a single ambient temperature swing that shifts both a pressure sensor and a temperature sensor, those terms are correlated and cannot simply be squared and added without accounting for the correlation. In practice, engineers treat the major contributors as independent when there is no shared mechanism, and handle known common-mode effects separately.
Consider a natural gas custody point measured with an orifice or ultrasonic meter feeding a flow computer. The uncertainty of the reported volume draws on the meter itself, the differential pressure or flow transmitter, the static pressure transmitter, the temperature transmitter and its element, and the gas composition or density input. Each of these carries a percentage uncertainty from its calibration and its datasheet.
To find TPE, you express every contributor as a percentage of the measured value, square each one, sum the squares, and take the square root of that sum. Suppose the flow element contributes one figure, the differential pressure transmitter another, and the static pressure and temperature transmitters smaller figures still. The RSS result will sit only modestly above the single largest term, which immediately tells you where accuracy is won or lost. Trimming a minor contributor barely moves TPE, while improving the dominant term moves it directly.
This is why TPE is the number that lands in a fiscal measurement report and in a custody transfer agreement, not the accuracy of any one instrument. A meter datasheet might promise a very tight flow-element figure, but the volume that changes hands and gets invoiced depends on the whole chain. Quoting only the meter's accuracy would understate the real uncertainty of the money-bearing number, and any auditor reviewing the measurement package will look for the combined figure.
In a cloud SCADA environment, the flow computer usually calculates the corrected volume in the field, and the platform historizes the flow, the differential pressure, the static pressure, the temperature, and the diagnostic values behind it. Because every input that feeds TPE is also a logged tag, the measurement report can be reproduced and audited from stored data rather than trusted on faith. When a discrepancy appears, having each contributing signal on the same trend lets you see which input misbehaved.
TPE also shapes how alarms and validation limits are set. If the combined uncertainty of a custody point is known, then a deviation between a metered volume and a downstream check volume can be judged against that band. A difference that sits inside the TPE window is measurement noise, not a leak or a theft; a difference well outside it is a real event worth investigating. Encoding the TPE figure alongside the point in the platform keeps that judgment consistent across shifts and operators.
For an operator running many wells and metering stations through Merobix, the practical payoff is that the accuracy claim attached to each custody point is documented, revisitable, and tied to real calibration records rather than living in a spreadsheet on someone's laptop. When a contract counterparty questions a volume, the combined-error story is already assembled from the logged inputs, which turns a dispute into a data review instead of an argument.
They are closely related and often used interchangeably in flow work. Total probable error is the specific root-sum-square combination of the individual error sources in a flow measurement chain, and it produces an uncertainty figure. Formal uncertainty analysis under the GUM framework uses similar mathematics but adds concepts like coverage factors and distributions, so TPE can be thought of as the flow-measurement dialect of combined uncertainty.
Because a straight sum assumes every error reaches its maximum at the same moment and in the same direction, which is statistically improbable for independent sources. Root sum square weights the large terms heavily and the small ones lightly, producing the realistic worst-likely combined error rather than an inflated absolute worst case. This is why RSS is the accepted method for independent contributors.
It depends on the metering type and operating point, but often the flow element or the differential pressure measurement carries the largest term, especially when a meter runs low in its range. Because RSS is driven by the largest squared term, identifying and improving that dominant contributor gives the biggest reduction in TPE, while chasing small terms yields almost nothing.
Merobix reads your field devices into a cloud SCADA - the real thing behind these terms, live in days from any browser.